Personalized Information Retrieval Systems: Enhancing User Interaction through Context-Aware Technologies

Abstract

The explosive growth of mobile Internet usage has led to an overwhelming proliferation of mobile applications. This phenomenon, commonly referred to as app overload, has resulted in significant challenges for users in identifying relevant and valuable content. Traditional recommendation systems often rely solely on historical usage data, neglecting critical contextual information, which led to suboptimal performance in dynamic usage environments. To address these challenges, this research introduces a Context-Aware Personalized Information Retrieval System (CAPIRS) that integrates a novel deep learning-based approach to enhance user interaction by capturing individual preferences across contextual scenarios. CAPIRS employs a dual-portrait modeling strategy, constructing comprehensive representations of both users and apps by incorporating attribute and contextual features. An Efficient Shark Search-driven Self-Attention-based Long Short-Term Memory (ESS-SLSTM) model is employed to accurately predict the probability of user engagement. Real-world data comprising user, app, and contextual interactions were collected from mobile platforms spanning diverse environments, temporal conditions, and demographic groups. Preprocessing involved data cleaning, deduplication and normalization of numerical features. Semantic features from textual app data were extracted using Term frequency-inverse document frequency (TF-IDF), while contextual attributes were embedded and structured into a tensor-based format for improved learning efficiency. Experimental results demonstrate that CAPIRS significantly outperforms traditional and benchmark methods in Precision@N, Recall, and Mean Average Precision (MAP) across personalized app recommendation tasks. This research underscores the importance of combining deep learning with context-aware technologies to develop more intelligent, adaptive, and user-centric information retrieval systems.